TBS

[PAKman]

What are the (tbs) files used for in the below example?
KNNNKN – 0.78 GB in 2 files (tbs | md5)
I know how to use the KNNNKN and the (md5) files.
But what is the use of the (tbs)?

Norm

[h.g.muller]

tbs files contain an overview of the statistic of the EGT, (i.e. the number of positions for each DTM), not? They are simple text files, of a few dozen lines.

[PAKman]

Hi, H G
What use is the information?
Does a chess playing program use the files?

Thanks for the replay.
Norm

[h.g.muller]

I don't think a chess-playing program needs them. (I have never used Nalimov tablebases in a Chess program, though, and do not have a single one on any of my computers, so you should not automatically believe that what I think is true...)

In fact the .tbs files are the only Nalimov files that are of any use to me, as they tell which endgames are won, and how easily. Furthermore, the statistics provides numbers against which I can test the results of my own tablebase generator. It was long ago I did this, though, and can't recall if the numbers given in the .tbs files for each DTM are properly corrected for symmetry, or if they were Nalimov-specific, so that I could only compare the maximum DTM. Nowadays I do use the numbers that Steven Edwards once posted on his website, (I dowloaded the whole set of statistics before he took it down again) I am sure these were properly symmetry corrected.

[PAKman]

Thanks H. G. for a great answer.
Norm

[guyhaw]

As John Nunn and others have pointed out, the raw EGT statistics can be most misleading. There are, e.g., a lot of positions where the defender will simply capture a piece that is 'en prise' or where some 'simple tactic' is available.

The endgame only becomes a 'typical position' if it is a deeper win/loss or indeed a draw.

g

[h.g.muller]

Indeed, and I am well aware of that. This is another reason why I switched from using Nalimov statistics to DTC: in DTC the tactical wtm positions with a hanging piece betray themselves by having DTC = 0 or 1. So to judge if an end-game is won I look if there is a sizable fraction of position that have DTC > 3. Some end-games (e.g. KRKB) have fairly long wins, and yet they form only a very minor fraction of the total number of positions. Usually they correspond to special cases, where the black King is in peril. So although I don't think you can describe such losses as tactical, they are certainly not very typical.

I also always print the number of wtm positions where black is in check: this gives an indication which fraction of the board i covered by white, and thus an estimate for the number of positions where another black piece is 'en prise' as well. The number of btm positions where white is in check then gives you an idea how large the probability is that a black piece is defended. Together this gives a reasonable stimate for the number of trivial wtm wins.

A case where I am in doubt if it should be called 'tactical' or not, is the 'chase'. In end-games with weak pieces where the opponent has a much stronger piece, there are usually positions where one of the weak pieces is cut off from protection by other friendly pieces, and can be chased by a stronger opponent piece until the board edge or corner prevents further escape, and the piece is lost. This happens, for instance, with Q vs N. (Sliders are usually too hard to catch.) In that case the chase does not last very long. But if poth hunter and pray get weaker, the chase can take longer. For instance, a King can uccesfully chase both Ferz and Wazir, so in end-games with the latter two pieces, there is always a very substantial fraction (~2-5%) of atypical results due to a Ferz or Wazir starting in a position where it is doomed.

Another issue that makes it unclear how I should classify end-games is the presence of fortress draws. E.g. KQKBN does seem generally won from the statistics, but with black: Ka1, Bb2, Nd4 there is no way K+Q could win. Even more owerful pieces than Q, such as Amazon or Lion cannot defeat this fortress. So a certain fraction of positions in a mostly-won end-game could get lost to fortress draws, and the overall winning fraction tells you how easy it is for black to build the fortress.